Insights / Professional services
The CRM failure rate everyone quotes is a misquote
The article
If you have been quoted for a CRM project in the last twenty years, someone has told you that most of them fail. Usually 70%. Sometimes 63%, or 50%, or "up to 90%."
We went looking for the source. What is there is instructive — not because the software industry is uniquely dishonest, but because the actual finding says something more useful than the myth, and it says it in your favour.
Where the number came from
The trail ends at a 2004 interview with Ed Thompson, then a Gartner analyst and co-author of Gartner's Eight Building Blocks for CRM Success.
The underlying research was one study, conducted at the back end of 2001, covering roughly 500 organisations in the US and Europe — all Gartner clients, all large.
The question asked was not "did your project fail." It was "did it meet expectations." 55% said it had not.
Thompson, in his own words:
"people took our statement: 'failed to meet expectations,' and they chopped the 'meet expectations' off it and just said, 'failure.' So, we were quoted by the press as saying 55 percent of projects are failing."
The actual distribution he described: about 5% absolute success, about 40% met expectations or were somewhat successful, about 40% somewhat successful but short of expectations, about 10% failed to meet expectations and were somewhat unsuccessful, and about 5% absolute failure.
Asked directly whether he agreed with the failure framing:
"No, I don't, because it's not absolute failure. And if you're looking at absolute failure, you're looking at 5 percent, maybe. Perhaps if you include the 'failed to meet expectations and somewhat unsuccessful' category that might add another 10 percent."
The 65–70% figure that outlived all of this was never a measurement at all. It was a Gartner Strategic Planning Assumption — a forecast that the rate would worsen. Thompson: "we never went back and did the same study to see if the 'failure to meet expectations' rate got worse."
The part that actually applies to you
Buried in that same interview is the finding nobody repeats, presumably because it does not frighten anyone into buying anything:
"They were Gartner clients. They were large organizations. They tend to have larger projects, and everything we've seen since shows that the smaller projects do much better. The mid-size organizations haven't had as much of a problem."
So the statistic used to scare a forty-person consultancy out of building anything was measured on enterprises, and its own author said mid-size organisations did better.
For what it is worth, the incoherence across the whole category makes the point on its own: published "CRM failure" rates range from 18% (AMR Research, 2005) to 70% (Butler Group, 2002), including Gartner 2001 at 50%+, the Economist Intelligence Unit at 56% in 2007, and Forrester at 47% in 2009. There is no agreed definition of failure. Which means any single number is a rhetorical device, not a measurement.
What is genuinely measurable: the tool count
Set failure rates aside. The real problem in a 10-to-100-person services firm is more mundane and better evidenced.
BetterCloud's 2026 report, surveying 525 IT and security professionals, found mid-market organisations averaged 164 applications, up 41% from 116 the year before. Also from that survey: 62% of IT leaders say manual work is actively preventing strategic projects, and 90% of organisations lack true cross-app orchestration — which BetterCloud describes bluntly, noting that what most orgs call automation "amounts to basic identity deprovisioning, not the kind of end-to-end workflow orchestration that actually eliminates manual work."
Okta's 2025 platform data put the average at 101 apps per company, the first time it crossed a hundred.
Both are vendors selling into this problem, and BetterCloud's own numbers are not consistent year to year (130 in 2023, 106 in 2025, up 11% in 2026). Cite one year. Do not draw a trend line through them.
Salesforce's State of Sales (fifth edition, 7,775 responses across 38 countries, fielded in 2022) found sales teams use an average of 10 tools to close a deal, 66% of reps say they are overwhelmed by the number, and 9 in 10 organisations plan to consolidate. Also from that research: reps spend just 28% of the week actually selling.
What switching between them costs
The best available measurement comes from a Harvard Business Review study (2022) that observed 137 users across 20 teams at three Fortune 500 companies, capturing around 3,200 days of work. It found:
- The average user toggled between applications nearly 1,200 times a day
- Reorienting after those switches consumed just under four hours a week
- That is about 9% of annual working time — "five working weeks"
An important caveat, because we would want it applied to us: the lead author is the founder and CTO of the company whose software collected the data, and the sample is Fortune 500 employees. It is not a measurement of a thirty-person consultancy. Use it as evidence that the coordination cost is real and large, not as your number.
A second data point, from Cornell's Ellis Idea Lab working with Qatalog (3,000 participants across the US and UK, 2021): people reported losing 59 minutes a day looking for information across apps, 44% said siloed tools made it hard to know whether work was being duplicated, and 48% said they were making mistakes because they could not track what was happening across systems.
That last one is the real cost, and it does not appear on any invoice.
What the services-firm benchmarks actually show
If you run a professional services firm, the number that matters more than app counts is utilisation — and it is worse at your size.
SPI Research's 2025 benchmark (403 firms, surveyed autumn 2024) breaks billable utilisation down by organisation size:
- Under 10 people: 64.3%
- 10–30 people: 65.4%
- 31–100 people: 69.2%
- 101–300 people: 72.4%
- All firms: 68.9%
SPI's stated optimal threshold is 75%. Small firms run six to ten points below it. Across the industry, utilisation fell from 73.2% in 2021 to 68.9% in 2024, on-time delivery fell from 80.2% to 73.4%, and EBITDA dropped to 9.8% — single digits for the first time.
The most interesting number in that report is not a headline: CRM-to-PSA integration sits at 51.3% among high performers versus 37.0% among everyone else — the largest relative integration gap SPI measured. Meanwhile commercial CRM adoption is near-universal in both groups (92.5% versus 81.6%).
Read those two together and you get the actual finding: having a CRM is not the differentiator. Connecting it to delivery is.
Which makes sense, because a generic CRM models a product sale and stops at closed-won — precisely where a services firm's economics begin. Staffing, utilisation, realisation, margin: all of it happens after the moment the CRM considers the story finished.
The failure modes we actually see
Buying more software. App counts rose 41% at mid-market while 62% still say manual work blocks strategic projects. Each new tool adds a boundary, and boundaries are where the coordination cost lives.
Automation duct tape. We will be honest that there is no rigorous independent research on no-code automation failure rates — everything circulating on that comes from agencies selling rebuilds. What we can say from our own delivery experience, labelled as experience rather than research: the dangerous failure mode is not the error, it is the absence of one. A trigger that stops firing produces no execution, therefore no error, therefore no alert. The dashboard is green because nothing ran.
Hiring an ops person to bridge the systems. This works, briefly. Then that person becomes the integration — and a human integration does not scale, cannot be audited, and leaves when they leave.
Treating consolidation as a project. BetterCloud's own data shows organisations consolidating for two years and then app counts rising again. Nothing structural changed about how systems connect, so the sprawl regrew.
The build/don't-build question
Here is where we are supposed to tell you to build custom software. Instead, three honest tests.
Do you have one process that genuinely does not fit the market, or ten that annoy you? Ten annoyances is a configuration problem. One genuine structural mismatch — usually where your delivery model differs from the standard — is worth building around.
Where is the handoff failing? SPI's data says sales-to-delivery integration is what separates high performers. If your CRM and your delivery system disagree about what was sold, that is the target. Not a better CRM.
Would you still want it if it were boring? The best answer we give is frequently "connect the two things you already own and stop." It is a smaller invoice for us and usually the correct engineering.
And if someone quotes you a failure rate to justify a purchase — or to justify not building — ask where the number came from. Twenty-five years of evidence says they will not know.
Common questions
Is it true that 70% of CRM projects fail?
No. The figure traces to a Gartner survey of about 500 large organisations conducted in late 2001, which asked whether projects met expectations — 55% said they had not. Gartner analyst Ed Thompson later said the press "chopped the meet expectations off it," and put absolute failure at around 5%, perhaps 15% including partial failures. The 65-70% number was a forecast Gartner never went back to validate.
Do smaller companies have worse software project outcomes?
The evidence points the other way. The Gartner analyst behind the original study noted the sample was large organisations with large projects, and that "everything we've seen since shows that the smaller projects do much better. The mid-size organizations haven't had as much of a problem." Scope, not company size, is the risk driver.
How many software tools does a mid-size company use?
BetterCloud's 2026 report, based on a survey of 525 IT and security professionals, found mid-market organisations averaged 164 applications — up 41% from 116 the prior year. Okta's 2025 platform data put the average across its customer base at 101. Both are vendor sources and their year-to-year figures are not internally consistent, so cite a single year rather than drawing a trend.
Bring us your messiest process. We’ll map it — free.
45 minutes with the engineers who’d build it. You leave with a build/don’t-build answer and an honest sense of cost — whether or not you work with us.
Book a working session →